Skip to main content

Install and load

Prerequisites​

  • DuckDB version 1.3 or newer (the CLI, the app, or any language client).

Install from the community repository​

Once the extension is registered in DuckDB's community extensions, installing is two statements:

INSTALL duckfn_statrs FROM community;
LOAD duckfn_statrs;

No extra flags needed — the community build is signed and matches your DuckDB version.

Install from a GitHub Release​

If the community route is not yet available for your platform, download the pre-built binary from the releases page. The asset is named duckfn_statrs-<platform>.duckdb_extension and has to be renamed to duckfn_statrs.duckdb_extension before it can be loaded: LOAD takes an extension name or a local path — it fetches no URLs (INSTALL '<url>' is the one that downloads) — and DuckDB takes the entry-point symbol from the file name, so the release name fails with did not contain function "duckfn_statrs-windows_amd64_init_c_api":

curl -LO https://github.com/shijianjs/duckfn-statrs/releases/latest/download/duckfn_statrs-windows_amd64.duckdb_extension
mv duckfn_statrs-windows_amd64.duckdb_extension duckfn_statrs.duckdb_extension
duckdb -unsigned -c "LOAD './duckfn_statrs.duckdb_extension';"
note

Released binaries are not signed by DuckDB's distribution key, so start the CLI with the -unsigned flag: duckdb -unsigned. The community route above does not have this limitation.

Verify the installation​

SELECT function_name, function_type
FROM duckdb_functions()
WHERE function_name LIKE 'sr_%'
ORDER BY function_name;
-- 511 rows

First queries​

Summary statistics are aggregates — pass a column, get one value per group:

SELECT g, sr_mean(x) AS mean, sr_std_dev(x) AS std_dev, sr_median(x) AS median
FROM (VALUES (1, 1.0), (1, 2.0), (1, 3.0), (2, 10.0), (2, 20.0), (2, NULL)) t(g, x)
GROUP BY g
ORDER BY g;

Distribution functions are scalars — evaluated row by row:

SELECT x, sr_normal_cdf(x, 0.0, 1.0) AS cdf, sr_normal_pdf(x, 0.0, 1.0) AS pdf
FROM (VALUES (-1.96::DOUBLE), (0.0), (1.96)) t(x);

Errors on invalid parameters are clear and specific:

SELECT sr_normal_pdf(0.0, 0.0, -1.0); -- error: sr_normal_pdf: Standard deviation is NaN, zero or less than zero

NULL behaviour​

  • NULL input rows never enter an aggregate (standard SQL convention).
  • When statrs cannot define a result (empty group, insufficient samples, out-of-range parameter), the return value is SQL NULL — never NAN.
  • A parameter that is present but invalid (e.g. std_dev <= 0) raises a query error.
-- statrs cannot define the sample variance of one value: the result is NULL, not 0.
SELECT grp, sr_variance(x) AS variance
FROM (VALUES ('two', 1.5::DOUBLE), ('two', 2.5), ('one', NULL::DOUBLE), ('one', 3.0)) t(grp, x)
GROUP BY grp
ORDER BY grp;

What to do next​